Rethinking Authorship under the Indian Copyright Regime
AUTHORED BY – KANISHKAA KUNDU
College: Sister Nivedita University, Kolkata, West Bengal
ABSTRACT:
Generative artificial intelligence tools that autonomously produce literary, artistic, and musical output have exposed a structural weakness in the Indian copyright framework: its insistence on a human, or at least a legally recognized, author. This article examines the doctrinal conflict between Sections 2(d) and 17 of the Copyright Act, 1957, which anchor authorship and ownership in a natural or juristic person, and the practical reality of machine-generated expression. Using the Ankit Sahni “RAGHAV/Suryast” episode before the Indian Copyright Office as the central illustration and drawing comparisons with the position in the United States, the United Kingdom, and the European Union, the article argues that India’s existing “computer-generated work” provision is doctrinally too thin to resolve the question of AI co-authorship. It concludes that legislative clarification rather than administrative improvisation is necessary to secure both creative incentive and legal certainty.
TO THE POINT:
Indian copyright law was drafted for a world in which every work traced back to an identifiable human mind. Generative AI systems disrupt that assumption. A user who types a prompt into a text-to-image or text-to-text model exercises far less creative control than a photographer composing a shot, yet the output can be indistinguishable from human-authored expression. The Copyright Act, 1957 offers only one textual hook for such situations the “computer-generated work” limb of the authorship definition and even that provision presumes a human “cause” the work, without addressing degrees of causation, joint AI-human contribution, or an AI system’s own claim to authorship.
The interests at stake are economic rather than ideological as well. Agencies for advertising, design bureaus, newsrooms, and software developers all depend more on the generative technology as part of their production chain, and the legal enforcement of copyrights for their creations will establish if they could be licensed, assigned, or protected from copyright infringement. Unless addressed, the uncertainty deters investment into artificial intelligence-based creative industries but exposes human creators at the same time to exploitation without retribution.
USE OF LEGAL JARGON:
Authorship refers to the person recognized in law as the originator of a work and the first holder of the bundle of exclusive rights that copyright confers. Originally, the threshold for protection under Section 13 of the Copyright Act, 1957, does not demand novelty but requires an independent, non-copied expression bearing a minimal degree of creativity or judgment.
Computer generated work is a concept derived from law in Britain, and it refers to any work done by a computer without the involvement of any human, in contrast to a computer assisted work where a human uses a computer merely as a medium of expression. Work-for-hire (or the employer-employee ownership rule under Section 17), and
“Skill and judgment” test— the standard Indian courts use to assess originality, discussed below together determine who, between a prompter, a developer, and a machine, may lay claim to a generative output. Finally, moral rights (the author’s personal right to attribution and integrity of the work, distinct from economic exploitation rights) remain conceptually difficult to extend to a non-human system.
THE PROOF:
The statutory starting point is Section 2(d) of the Copyright Act, 1957, which defines “author” separately for each category of work an artist for artistic works, a composer for musical works, and, critically, under sub-clause (vi), “the person who causes the work to be created” for literary, dramatic, musical or artistic works that are computer-generated. Section 17 then vests first ownership of copyright in that author, subject to statutory exceptions for commissioned and employment works. Read together, these provisions presuppose a legal person natural or juristic standing behind the creative act, a premise consistent with Article 2 of the Berne Convention, to which India is a signatory, and which frames copyright protection around “literary and artistic works” without contemplating non-human creators.
On originality, Indian courts have charted a middle path between the low threshold “sweat of the brow” doctrine and the demanding requirement of novelty. In Eastern Book Co. v. D.B. Modak, the Supreme Court held that a work must display a modest “flavor of minimum creativity” beyond mere Labour, echoing the Canadian “skill and judgment” standard. In R.G. Anand v. Delux Films, the Court reaffirmed that copyright subsists in the expression of an idea and not the idea itself. Both tests were formulated for human creative choices selection, arrangement, and expression and translate awkwardly to a large language or diffusion model whose “choices” are statistical inferences drawn from training data rather than conscious judgment.
Comparative law illustrates the range of possible responses. The United Kingdom’s Copyright, Designs and Patents Act 1988 expressly contemplates computer-generated works with no human author, assigning ownership of works to “the person who makes the necessary arrangements for the production of the work,” while providing only fifty years of copyright protection, an unusually short period considering the default copyright duration of life plus sixty years.
The United States has taken the opposite, stricter route. The Copyright Office’s Compendium of U.S. Copyright Office Practices denies registration to works “produced by a machine or mere mechanical process” absent human creative input, a position the D.C. Circuit endorsed in Thaler v. Perlmutter. The European Union, meanwhile, has so far declined to create a sui generis authorship category for AI output, instead regulating the upstream problem through the AI Act’s training-data transparency obligations, which require general-purpose AI providers to disclose a summary of copyrighted material used in training an approach aimed at protecting human authors whose works feed these systems rather than at resolving downstream ownership of AI output.
India has neither codified a UK-style computer-generated-works category nor adopted the US bright-line human-authorship rule; it has instead allowed the question to be litigated by administrative accident, and a Parliamentary Standing Committee has already flagged the need for legislative review. The absence of guidance also intersects with the fair-dealing exceptions under Section 52, since the legality of using copyrighted material to train Indian-facing AI models remains untested, a question now squarely before the Delhi High Court.
CASE LAWS:
- The RAGHAV/Suryast episode
In 2020, Ankit Sahni sought registration for Suryast, an image generated by his AI tool RAGHAV from a photograph he had taken, styled after Vincent van Gogh’s The Starry Night. An application naming RAGHAV as sole author was refused, but a subsequent application naming Sahni and RAGHAV as co-authors was registered in November 2020 reportedly the first instance globally of an AI system being recognized as a copyright co-author. The Copyright Office subsequently issued a withdrawal notice questioning RAGHAV’s status as an “author” under Section 2(d)(iii) and (vi), and the matter remains formally unresolved: the registration has neither been judicially set aside nor conclusively reaffirmed. The episode is significant precisely because it is not a reasoned judicial precedent but an unresolved administrative signal exposing the absence of settled law.
- Thaler v. Perlmutter
Stephen Thaler sought US copyright registration for an image generated autonomously by his AI system, DABUS, listing the machine as sole author. The Copyright Office refused registration, and the D.C. Circuit affirmed in 2025 that the Copyright Act’s use of terms such as “written” and the constitutional reference to “authors” presuppose human agency, foreclosing copyright in wholly machine-generated output.
- Naruto v. Slater
Although not an AI case, the Ninth Circuit’s ruling that a macaque could not hold copyright in a self-taken photograph is regularly invoked in AI-authorship debates for its underlying principle: statutory authorship is limited to persons recognized by law, a limitation courts have since extended by analogy to autonomous machine output.
- Zarya of the Dawn
The US Copyright Office Review Board’s 2023 decision on a Midjourney-assisted comic book drew a workable line for hybrid works: copyright was granted for the human-authored text and the creative arrangement of images and text but denied for the individual AI-generated images themselves, since the prompts did not give the author sufficiently predictable control over the final visual output.
- ANI Media v. OpenAI
Pending before the Delhi High Court, this suit alleges that OpenAI’s large language models were trained on ANI’s copyrighted news content without licence. The outcome will shape whether India’s fair-dealing exception under Section 52 extends to text-and-data mining for AI training, a question distinct from, but closely linked to, the authorship debate.
CONCLUSION:
The RAGHAV episode did not create India’s AI-authorship problem; it merely made an existing statutory gap visible. Section 2(d)(vi) was drafted for early computer-assisted tools spreadsheets, weather models, database compilers in which a human plainly “caused” the output through direct instruction. Generative AI systems trained on billions of pre-existing works and capable of producing unpredictable, stylistically rich content strain that provision past its intended scope. Administrative improvisation, whether registering and then withdrawing a co-authorship claim, or fast-tracking an application without a reasoned order, cannot substitute for legislative clarity.
This author’s view is that India should legislate a graduated framework: works produced with substantial human creative control (prompting, curation, editing) should remain protectable in the human user or commissioning entity, consistent with existing skill-and-judgment jurisprudence; wholly autonomous machine output, with negligible human contribution, should fall outside copyright altogether, as the US position holds, rather than being awarded a diminished UK-style term to an ambiguous “person.” Simultaneously, Parliament should address the upstream question of training-data use through a calibrated text-and-data-mining exception, informed by both the EU’s transparency model and the outcome of ANI v. OpenAI, so that the rights of human authors whose works train these systems are not sacrificed to expedience. Until such reform occurs, applicants, licensees, and litigants in India will continue to operate in the shadow of an unresolved registration rather than a settled rule of law.
FAQs:
Q1. Can an artificial intelligence system be recognized as an “author” under Indian copyright law?
No. Section 2(d) of the Copyright Act, 1957 defines authorship by reference to categories of human or juristic persons, and even the computer-generated-work limb attributes authorship to the “person who causes the work to be created,” not to the machine itself.
Q2. What is the current status of the RAGHAV/Suryast registration?
It is still an open-ended administrative issue: the Copyright Office granted co-authorship of the work to both Ankit Sahni and RAGHAV in 2020 and withdrew its decision in 2021 regarding the legal validity of RAGHAV, without any reasoned conclusion to this effect.
Q3. Who holds the copyright in the output created by use of a tool like ChatGPT, Midjourney, or another similar model in India?
The Act does not say so expressly. On present principles, ownership most plausibly vests in the human user only where that user exercises sufficient creative selection, arrangement, or editorial control over the output; purely mechanical prompting is unlikely to satisfy the originality threshold.
Q4. Does India have a UK-style “computer-generated works” provision with no human author?
Not in the same form. The UK’s Copyright, Designs and Patents Act 1988 expressly allow for a computer-generated work with no human author and shortens its protection term; India’s Section 2(d)(vi) still requires a human “cause,” making the position narrower and less settled.
Q5. Why does the ANI v. OpenAI litigation matter for AI copyright policy in India?
It will test whether using copyrighted material to train AI models falls within the fair-dealing exceptions of Section 52, a threshold question that determines the legality of the training process underlying most generative AI output.
